It has long been known that presenting a face upside-down dramatically affects its recognition, but Yin (1969) was the first to show that inversion affects the recognition of faces more than for other object categories. Since then, researchers have conjectured about the sources of this 'face inversion effect'. While faces are recognized both on the basis of local features and their spatial relationships, there is a fairly large body of evidence that inversion affects more these spatial relationships, or face configuration (see our review on this issue: Rossion & Gauthier, 2002; as well as Maurer et al., 2002).
We have recenlty shown that while inversion affects equally well the processing of local features and of horizontal distances between features, the perception of vertical distances is spectacularly impaired (Goffaux & Rossion, 2007).
This does not appear to be related to the axis of rotation because vertical relationships are also more affected when faces are flipped horizontally (90 degrees).
In this line of research we have essentially investigated how inversion affected the early stages of face processing using ERPs.
In parallel with the findings of Yin, we have shown that inversion does not only increase and delay the N170 (Bentin et al., 1996; Rossion et al., 1999 - see related work on the N170), but that these effects are specific for faces or larger than for other objects (Rossion et al., 2000). In fact, the N170 delay can be found for pictures of highly familiar objects such as cars or words (Rossion et al., 2003) or the same novel objects after expertise training (Rossion et al., 2002; Busey & Vanderkolk, 2005), but it is larger for faces.
The fact that face inversion affects the N170 so strongly is indirect evidence that face configuration is extracted as early as 130 ms (onset of N170).
We have also studied face inversion in fMRI and shown that the source of the face inversion effect may lie in visual areas responding preferentially to faces ('FFA' and 'OFA'). We used an adaptation paradigm to show that there is larger adaptation (lower signal) in these areas when different facial identities are presented upside-down than when they are upright (Mazard et al., 2006). This differential adaptation rate accounts for a large part in the overall smaller response found in these regions for inverted faces.
Currently, we are running behavioural studies aimed at clarifying further the kind of spatial relationships are more impaired following upside-down face inversion. Our ERP experiments aim at clarifying the nature of amplitude and latency effects of inversion on the N170, something that has not been clarified yet in the literature.
Van Belle, G., de Graef, P., Verfaillie, K., Rossion, B., Lefèvre, P. (2010). Face inversion impairs holistic perception: Evidence from gaze-contingent stimulation. Journal of Vision. May 1;10. pii: 10.5.10. doi: 10.1167/10.5.10. [PDF]